This week Zeynep Tufekci, an assistant professor in the Department of Sociology at University of North Carolina, Chapel Hill*, talks with us about the implications of algorithmically filtering social media feeds.
In this case, the fact that the brain – just like the internet – is a network with “small world properties” helps. Every pixel in the brain and every Internet page can be seen as a hub in this network. The hubs can be directly connected to each other just as two Internet pages can be linked.
With eigenvector centrality, the hubs are assessed based on the type and quality of their connections to other hubs. On the one hand, it is important how many connections a particular node has, and on the other, the connections of the neighbouring nodes are also significant. Search engines like Google use this principle, meaning that Internet sites linked to frequently visited sites, like Wikipedia, for example, appear higher in results than web pages which don’t have good connections.
“The advantages of analyzing fMRI results with eigenvector centrality are obvious,” says Gabriele Lohmann from the Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig. The method views the connections of the brain regions collectively and is computationally efficient. Therefore, it is ideal for detecting brain activity reflecting the states that subjects are in.